Healthcare AI Daily Briefing

Medical Imaging

Deep learning segmentation model developed to improve emphysema quantification versus LAA-950.

Journal article

A deep learning segmentation algorithm was created using retrospective virtual and clinical CT datasets, with virtual data from 20 human models and clinical data from multi-center cohorts of 101, 23, and 1159 patients, to enhance accuracy and robustness of emphysema measurement compared to the LAA-950 biomarker. [1]

Research & Evidence

LLM after-visit summaries scored higher on understandability and actionability than physician versions.

Journal article

In a blinded retrospective study of 50 discharged adults, GPT-4 and Gemma summaries achieved higher PEMAT understandability (85.5%, 87.5% vs 66.1%) and actionability (70.9%, 74.1% vs 56.7%; all p<.001) and were rated better on the AVSrubric for clarity, comprehensibility, empathy and lower potential for harm. [2]

Industry & Products

AI‑driven tool tracks physician tasks per contract in Toledo pilot.

Podcast from Healthcare AI Pioneers

A speaker described an AI‑driven product that lets physician leaders log tasks aligned with their contracts, originating in the Toledo market with input from physicians and the compliance team who conceptualized it, and helps decrease the risk associated with administering those contracts. [3]

Industry & Products

Oura files for IPO with $1.2B nine‑month revenue and 5M members.

Oura’s S‑1 shows $1.2B revenue for the nine months ended June 30, up 74% YoY, and $61M net income on 55% gross margins. It has 5M paid members wearing the ring 23 h/day, 94% activation‑to‑membership conversion and 85% annual retention, plus 42B hours of biometric data and 1,200+ integration partners. [4]

Clinical Practice

Speaker used AI to retrieve U.S. ALS organization contacts in about fifteen minutes.

Podcast from The Business of AI in Healthcare

A speaker said they used AI to get names, phone numbers and details of all U.S. ALS organizations, a task that would have taken weeks manually, in about fifteen minutes. [5]